FunClip is a local Gradio-based video clipping tool that uses FunASR for transcription and supports LLM-assisted clipping with multiple ASR models. It offers multi-segment clipping, subtitle generation, and optional speaker recognition features, with releases updating Gradio compatibility and transcript handling.
Collecting history — the radar snapshots this repo daily. The trend line appears after 3 days of data (1 so far).
What it is
FunClip is a fully open-source, locally deployed automated video clipping tool. It leverages Alibaba TONGYI open-source FunASR Paraformer models for speech recognition on videos, enabling users to select text segments or speakers from the recognition results to clip corresponding video segments. It provides multi-segment clipping and returns full video SRT subtitles and target segment SRT subtitles. It can be deployed via a local Gradio UI or on a server.
How it works
FunClip uses FunASR-based ASR (including Paraformer options) to transcribe video content. It supports hotword customization and speaker recognition (CAM++), enabling clipping by speaker ID. An optional TwelveLabs Pegasus integration allows video-visual reasoning for segment selection. Clipping and subtitle generation are driven through a Gradio UI, with options to generate and export subtitles for clipped segments.
Getting started
Install instructions include cloning the repo and installing Python requirements, with a note on Gradio 4 compatibility and required FunASR versions. Commands shown in the README include:
# clone funclip repo
git clone https://github.com/modelscope/FunClip.git
cd FunClip
# install Python requirments
pip install -r ./requirements.txt
To run locally via Gradio service, the usage example:
python funclip/launch.py
Additional notes mention updating to FunASR >= 1.3.29 if needed and container startup hints using --listen and --share.
Recent releases
- v2.1.1 FunClip v2.1.1 (2026-08-03): patch release for fresh Gradio 4 installations, container startup privacy, transcript matching case-insensitivity, and new provider routes.
- v2.1.0 FunClip v2.1.0 (2026-07-24): first versioned release, stable rollback point for current local video-clipping application.
Traction
Stars: 6104
Behind the repo
Not included in the provided facts.
Caveats
License: MIT. Model weights are downloaded separately and are governed by model pages. FunClip requires FunASR paths (funasr>=1.3.29) for FunClip’s Fun-ASR-Nano, SenseVoice, and subtitle compatibility.